AI tools deployed in banking oversight

The Bank of France has implemented five artificial intelligence applications in 2025 and early 2026, focusing on banking supervision, anti-money laundering efforts, and support for individuals. These tools are part of a broader initiative that began before the widespread use of conversational AI, with over sixty experiments in artificial intelligence and data science since 2015.

First AI assistant aids supervisors

One of the tools assists bank supervisors by responding to queries and generating summaries based on data relevant to their work. This application has reportedly saved each user about two hours per week, according to a report published in October 2026.

Other assistants handle internal and customer inquiries

Two additional AI assistants manage document-related tasks. One addresses internal human resources questions, while the other supports customer service agents dealing with individual requests, especially those related to over-indebtedness and the right to banking. Accuracy rates for these tools are reported to be above 90%.

AI improves money laundering detection

The fourth application helps identify suspicious transactions in the fight against money laundering and terrorist financing. Initial results suggest a significant decrease in false alerts generated automatically, improving the efficiency of the process.

Mapping financial products with AI

The fifth tool maps structured financial products to help analyze exposures and generate descriptive statistics. The institution claims its analyses are reliable at a 95% level, aiding in more accurate financial reporting and decision-making.

AI used for anomaly detection and data analysis

Algorithms are also being used to detect anomalies in bank statements, merge databases, and analyze public perceptions of inflation. An experiment in 2026 focused on reconstructing business group structures from diverse data sources, enhancing the accuracy of financial oversight.

Natural language processing aids regulatory reviews

For the Prudential and Resolution Authority (ACPR), natural language processing helps review regulatory documents more efficiently. These technologies can identify unusual patterns and guide inspections toward complex cases that require deeper analysis.

AI integration in the Bank's strategic plan

The Bank of France has included AI in its 2026-2028 strategic plan, aiming to transform internal processes and expand AI tool access for staff. However, the results so far focus only on identified applications and do not provide a full assessment of productivity improvements.

Operational risks in AI deployment

Banking applications involving AI deal with sensitive financial and personal data, which introduces risks such as factual errors, biases, performance decline with changing data, difficulty in explaining AI decisions, and potential leaks of confidential information. The reliability of these tools depends heavily on the quality of input data and human oversight.

Internal AI advisory group established

An internal advisory group on AI, formed in 2025, brings together legal, computer science, ethical, and data protection experts. It has provided about twenty opinions in one year on existing AI projects, ensuring compliance with both internal and European AI regulations.

Oversight aligns with European AI regulations

The Bank of France notes that its AI oversight aligns with the European Union's artificial intelligence framework, which is being implemented gradually based on risk levels. The ACPR is adapting its audit methods to address AI systems used in financial assessments, such as evaluating solvency or insurance pricing.

Future assessments of AI performance

As AI tools become more common in regulatory and public service roles, the Bank of France plans to evaluate their long-term effectiveness, including the quality of responses, the reduction of unnecessary alerts, time saved, and the ability of staff to verify AI-generated results.